HGAPS/Asian identity and cultural context on eating psychopathology

Examining the impact of Asian identity and cultural context on eating psychopathology

Hannah Kim

The University of North Carolina at Chapel Hill

Author Note

This project was supported by the Tom and Elizabeth Long Excellence Fund for Honors administered by Honors Carolina, the Lindquist Undergraduate Research Award, and the NC TraCS Institute (CTSA—UL1TR002489).

Literature search
Eating disorders have the highest mortality rate of all psychiatric illnesses (Arcelus, 2011). While America and many Western countries have experienced a cultural shift towards acceptance of body diversity and promotion of body positivity, Asian countries and especially East Asian countries continue to hold strict expectations of thinness. Despite increasing globalization, more cross-cultural contact than ever, and a growing Asian American population, available research on the mental health of Asian Americans is extremely limited. Coupled with the fact that Korea consistently has one of the highest suicide rates of the developed world and that even unplanned suicide attempts are highly correlated with pre-existing mental illness (Jeon et al., 2010), identifying whether disordered eating is of high concern among Asian and Asian American populations, which risk factors are of particular concern, and whether our current measures behave similarly for Asian and non-Asian populations can provide us a more accurate picture of a largely under-researched concern for this population.

Background information on eating disorders
Eating disorders (EDs) refers to the category of psychiatric illnesses in which people experience severe disturbances in their lives related to eating and food, which can present as behaviors, thoughts, and emotions. Defining EDs has been historically inconsistent even within the field of psychiatry as reflected in the multiple changes to DSM criteria between each edition, and to this day there are continued disagreements in definitions, such as the one over the incorporation of a weight or BMI threshold criterion for anorexia nervosa.

The term “eating psychopathology” or disordered eating serves to describe a continuum of risky attitudes and behaviors regarding eating, food, and body image that may or may not be part of a formally diagnosed eating disorder. Given the underutilization of mental health resources (Leong et al., 2007) as well as poor overall awareness of EDs, this study primarily focuses on broad eating psychopathology as opposed to specific diagnoses. In addition to capturing a population without formal diagnoses, eating psychopathology captures those who are likely at risk of developing later EDs.

Risk factors
Keery et al. (2004) theorize that external sociocultural influences contribute to eating psychopathology along with an individual’s tendency to compare themselves to others and to internalize those external influences through what they call the Tripartite Influence Model of body dissatisfaction. This model suggests that peers, family, and the media can all influence the way in which individuals engage in comparison and internalization, meaning that people in different social environments with different cultural norms may have different levels of risk of developing EDs.

Perfectionism has additionally been suggested to be a transdiagnostic risk factor for eating disorders, especially self-oriented perfectionism stemming from self-imposed standards (Leong et al., 2007). Furthermore, it is possible that perfectionism could be a maintenance factor in eating pathology (Stice, 2002) and the high levels of perfectionism associated with individuals with ED diagnoses and can endure even after recovery (Bardone-Cone et al., 2007).

Necessity of representation
The primary way in which Americans and Western countries have been exposed to Korean culture in particular is through K-pop, which also serves as a foil for the harshly restrictive attitudes towards weight held by much of the Korean population. Male and female idols alike are praised for their almost otherworldly beauty, though their appearances are often augmented with plastic surgery, and inextricably linked to that beauty is their thinness. Their appearances often come at a cost, and there are numerous accounts of abuse within the industry on part of management corporations. K-pop singer IU’s dieting plan shocked mainstream American media two years ago and even led to YouTubers trying the diet and calling out its infeasibility and dangers. On this plan, IU only eats one apple for breakfast, 1 or 2 sweet potatoes for lunch, and one protein drink for dinner, which totals only around 300 calories per day. She later admitted to a diagnosis of bulimia nervosa, and while this diagnosis was met with heartfelt support from some fans, many others made hateful comments telling her that she should lose more weight, that she is too weak-willed, and others telling her to simply end her own life. While celebrities like IU are under the most scrutiny by the Korean public with respect to appearance, this scrutiny encourages continued expectations of thinness, which in turn influences the ideal body image of the rest of the population, especially among impressionable media-consuming adolescents. This body image has even been quantified in popular culture – notably the 50-kg standard. Women over 50 kg (110 lbs) at any height are considered fat. DSM-5 criteria states that having a BMI under 17.5 can qualify an individual for anorexia, so any woman over 167 cm or 5’6” weighing 50 kg could meet diagnostic criteria for AN, and any woman above 5’5” weighing 50 kg would be considered underweight (BMI<18.5) (American Psychiatric Association, 2013).

Although these myths and media representations are serious, they alone do not communicate the severity of body image standards as well as empirical data. As indicated OECD Obesity Update of 2017, Japan and Korea respectively have the lowest and second lowest rates of obesity (3.7% and 5.3%) (Obesity Update—OECD, 2017). For the sake of comparison, this rate is the highest in the United States at 38.2%. Despite this indication that East Asian countries as a whole weigh less and are thinner on average than their Western counterparts, a national survey conducted by the Korean Ministry of Gender Equality and Family in 2016 indicates that over 60% of Korean women and 41% of men were either already on a diet or had a desire to start (Ministry of Gender Equality and Family, 2017). Perhaps of even greater concern is that 72% of girls under the age of 18 and 36% of boys in Korea believe they need to lose weight. This self-reported desire to lose weight that is ingrained well before adulthood mirrors the clothing market as well - the smallest size that is considered plus-size in Korea is 66, which is equivalent to a US size 6. A size 6 is typically qualified as small or small/medium in America, where the dress size of the average American woman lies somewhere between a size 16 and 18.

Impacts of migration and globalization
As the world rapidly globalizes, the differentiations between cultures have become more fluid, and people are more likely than ever to receive social influence from cultures around the world. People gain more exposure through travel, studies abroad, and social media. Additionally, race is socially constructed and racial categories do not capture the scope of differential cultural and social experiences that could influence eating disorder etiology. For example, customs and social norms are vastly different between China and Korea alone, let alone between East Asia and Southeast Asia or South Asia. Despite the wide variety of sociocultural influences across Asia, Asians are often treated as a monolithic racial category in American social sciences, which is unsurprising considering that Asians still only account for a small portion of the population. The Asian American population has grown more than tenfold from less than 1 million in 1960 to over 10 million in 2000. Additionally, 70.7% of Asian Americans are foreign-born, indicating that the majority of Asian Americans have lived in another country at some point in their lives (Office of the Surgeon General (US) et al., 2001). Given my limitations for this project, I will focus on primarily East Asian characteristics. However, more research needs to be done that is more inclusive of the diversity of backgrounds that Asian Americans possess.

Collectivism
Collectivism is generally defined as the practice of putting the interests of the group or community over any given individual, including oneself. This can best be contrasted to the American mindset of individualism, independence, and the bootstraps-mentality that encourages people to do what is best for themselves. Asian cultures generally tend to promote more collectivism than individualism. It seems that collectivism plays a significant role in the identity development of Asian Americans as well. When Yeh & Huang (1996) researched the ethnic identity development of Asian American college students, they found that their sense of identity was heavily influenced by their relationships with others as well as external forces. In particular, the avoidance of shame was found to be a significant motivating factor in determining their personal identities, which may play a role in the observed trends for perfectionism and mental health stigma in Asian American communities.

Perfectionism
It has been documented that Asians tend to endorse more perfectionism than Whites, especially the dimensions of parental expectations and parental criticisms (Yoon & Lau, 2008). This suggests that Asian cultures are more conducive to the development of perfectionistic characteristics, which may in turn adversely increase the risk of developing EDs given what we know about the role of perfectionism as a transdiagnostic risk factor.

Mental health stigma
As stated previously, Asians are less likely to utilize mental health services even when they are readily available (Leong et al., 2007). This is unlikely to solely be due to lower rates of mental illnesses in Asians. In fact, a key factor in preventing people who would benefit from such resources from reaching out may be mental health stigma. Compared to other racial groups, Asians experience greater stigma around mental illnesses (Kramer et al., 2002). Additionally, when Asian patients to express psychological distress, they are more likely to express them as physical complaints than emotional. The adherence to strong family and societal units seems to play a role in contributing to the self-perception of shame in admitting weakness in mental fortitude. Whether this stigma is more due to internal or external factors, it contributes to the decreased likelihood that Asian Americans seek professional help for mental health concerns.

Identity stress
Being placed in an unfamiliar environment with people who do not speak, behave, think, or even look like you can create psychological distress, especially identity stress related to minority status. An individual experiencing minority stress may be more likely to engage in performative behavior either not to harm social perceptions of other people in their group or to dispel stereotypes through stereotype threat (Kramer et al., 2002). A person who is part of a minority group is also more likely to experience imposter syndrome, which is even more strongly correlated with mental health problems than minority status stress (Cokley et al., 2013). Combined with the observation that most Asian Americans are immigrants or descendants of immigrants, minority stress that is caused by being a cultural outsider must be addressed as a potential risk for developing any mental health issues, including EDs.

Current status of research
The question of whether Asian Americans truly experience EDs, disordered eating, and other mental health issues at higher, lower, or approximately similar rates to other racial groups remains relatively unresolved. A review of existing literature by Yu et al. (2019) found that the research is mixed on the prevalence of EDs in Asian Americans, with some papers reporting that it is lower than in non-Asians while others report it as higher. They hypothesized that these mixed results may reflect the relatively low rate of help-seeking behavior in this population. A study of broader psychological distress among college students found that Asian American college students and adolescents experience greater psychological distress than their European American counterparts (Siegel et al., 1998). It is also possible that being Asian American in and of itself is not as important as how acculturated an individual is, and that studies suggesting more or less distress in this group are picking up on acculturative distress in this population that is overwhelmingly composed of immigrants. Wang and Mallinckrodt (2006) found that Chinese international students that reported a greater degree of acculturation also reported less psychological distress, which ties back to minority stress theory. In order to explore as many of these factors as possible, our study consists of a survey to administered to Asians and non-Asians alike, measuring disordered eating symptoms, Asian acculturation and identity, Asian-American values including collectivism and conformity to norms, multiple dimensions of perfectionism, and attitudes towards seeking professional help for mental health concerns. Our hope is that our results may add to the modest body of research on disordered eating in Asian Americans, and assist in illuminating the truth about how a multitude of sociocultural factors may interact with disordered eating.

Participants
N=361 participants ages 18 to 83 completed the survey for this study. Although disordered eating attitudes and behaviors can certainly develop in adolescence or even childhood, many of the measures included in this study are only validated for use with adult participants. For this reason, the lower limit on age is 18. No upper limit was defined in order to capture as many interested participants as possible, and potential cohort or age effects could convey a different story about societal expectations and generational culture shifts. It is also common within ED research to limit the population of interest to females since the rate of ED diagnosis tends to be higher in women. However, given that perceptions and expectations of body image and weight seem more restrictive for Asian men than others albeit less so than Asian women, men were also surveyed to determine whether there is a significant difference in eating psychopathology between Asian and non-Asian men.

The North Carolina Translational Clinical Science (NC TraCS) institute enables UNC researchers to enroll with Join the Conquest and ResearchMatch, which are online platforms funded by the NIH that match people interested in participating in clinical research with studies of interest. While Join the Conquest is primarily comprised of people who have received care from the UNC Healthcare system, ResearchMatch is a national platform with over 5,000 registered volunteers who are Asian by race. I used both of these platforms to recruit participants, in addition to mass emails to listservs or organizations and word-of-mouth. Since the population of interest is those who identify their race as Asian, the possible pool of participants is already severely limited since Asians comprise the smallest standard race group. To address this issue, Asians were oversampled by targeting Asian cultural organizations and specifying Asian interest within recruitment materials in ResearchMatch.

Of the 361 participants who completed the survey, 61.5% indicated that they found the survey through ResearchMatch, making it the most effective method of recruitment for this study. The second most popular participation source was through email (28.5%), which encompasses listserv announcements and forwarded emails from peers. In comparison, Research for Me (1.0%), the new human research platform that was rolled out in January 2020 by the NC TraCS Institute, and flyers (0.5%) were the least common platforms on which participants found this study. A likely explanation for why ResearchMatch was particularly effective is that it is researcher-initiated, as opposed to Research for Me and other platforms that require participants to find the listing before expressing interest. It should be noted that about a week after the start of recruitment, I made the decision to primarily focus my efforts on ResearchMatch and listserv announcements based on the return on time investments. ResearchMatch additionally allowed me to choose to send emails to participants who fit specific demographic criteria, enabling me to effectively oversample Asian participants by selecting that demographic category as my target for initial emails. After the system sends the initial email to up to 5000 randomly selected participants who fit the criteria a researcher have set, they will opt-in and indicate interest, releasing their contact information to the researcher’s dashboard for them to follow up and send them the link to their survey. This greatly diminishes the burden of effort on part of both the researcher and participant, and allowed me to receive complete responses from hundreds of participants in about three weeks.

Measures
The survey, which is presented under the secondary title of the Asian Cultural Identity and Eating Attitudes (ACIEA) Study, was administered through Qualtrics under the license of UNC Chapel Hill. This survey, with the exception with initial demographic and history questions, is entirely composed of other empirically developed questionnaires and scales that are intended to measure the constructs of interest.

Demographics
Standard demographic questions regarding age, gender, and race and ethnicity are included for ease of comparison to other research, replicability, as well as to analyze potential interaction effects with the key variables of interest. An additional question is asked to distinguish whether those who are Asian by race are East Asian or not since the central constructs of this study relate to East Asian populations. We also ask about educational attainment, US region of residence, and Asian region of background. Additional questions ask about relevant mental health treatment history – specifically, history of ED diagnoses, any other mental health diagnoses, and whether the participant has ever thought professional help for mental health reasons.

Eating Disorder Examination Questionnaire (EDE-Q)
The EDE-Q (version 6.0) is a 28-item measure derived from the Eating Disorder Examination (Fairburn & Beglin, 2011). The Global score is measured as the average of four subscales – Eating Concern, Weight Concern, Shape Concern, and Restraint. The 22 7-point Likert-type items are used to calculate these scores, and the remaining 6 items provide qualitative and frequency data about disordered eating behaviors. Generally, a score of 4 or higher is indicative of the clinical range, and Mond et al. (2004) suggest a clinically significant cut-off of 2.3 on the Global Score.

Suinn-Lew Asian Self-Identity Acculuration Scale (SL-ASIA)
The SL-ASIA measures acculturation in the direction from Asian to Western society through a 26-item scale (Suinn et al., 1992). These items ask participants to say how much they agree with a statement, what their preferences are, as well as what their cultural background is to produce an aggregate score based off of the initial 21 items. The last 5 items give orthogonal scores on whether a person is more Asian-identified, Western-identified, bicultural, or neither in terms of personal values, behavioral competencies, and self-identity. Many of these items are not relevant to participants who do not identify as Asian, but will be presented to everyone for the sake of consistency and to potentially provide a baseline for analyses.

Asian American Values Scale – Multidimensional (AAVS)
The AAVS is a 42-item scale that measures values associated with Asian Americans (Kim et al., 2005). The scale is divided into 5 components: collectivism, conformity to norms, emotional self-control, family recognition through achievement, and humility. For this study, we are primarily interested in the constructs of collectivism, conformity to norms, and emotional self-control as they have the most relevance to disordered eating, and for the sake of brevity we only asked participants to answer the first 3 sections, totaling 22 items. The question asks participants to rate how much they agree with a given statement that aligns with a certain value on a 7-point scale. This measure was specifically developed and validated for use in Asian Americans, but the questions are worded such that meaningful data can be collected from non-Asian identifying participants as well.

Frost Multidimensional Perfectionism Scale (FMPS)
The FMPS is a 35-item multidimensional self-report scale of perfectionism (Frost et al., 1990). The scale is composed of six subscales: Concern over Mistakes, Personal Standards, Parental Expectations, Parental Criticism, Doubts about Actions, and Organization. The total score is based on the 29 items that exclude Organization, which was only found to be loosely related to the other five subscales by Frost et al. (1990).

Attitudes Toward Seeking Professional Psychological Help Short Form (ATSPPH-SF)
The ATSPPH-SF is a 10-item scale used to measure general attitudes towards seeking professional psychological help on a 4-point Likert-type scale where a higher score indicates more positive attitudes (Fischer & Farina, 1995). A factor analysis revealed the three dimensions of openness to seeking professional help, value in seeking professional help, and preference to cope on one’s own (Picco et al., 2016). In this study, this measure is being used to estimate participants’ mental health stigma as well as likelihood of seeking help.

Procedures
The manner in which participants gain access to the survey link differ depending on which recruitment method was used to reach them. They either access the survey link directly from a webpage, as will be the case with Join the Conquest, or more likely via email, either directly from myself as expected with ResearchMatch or over a mass email such as listservs. They may also manually enter the web link themselves if they see it on a flyer, or use a QR code such as the ones on physical flyers.

The initial conservative estimate of the expected length to complete the survey was one hour if participants take 20 to 30 seconds per question. Given that the majority of the survey use likert-type scales and based off the feedback of friends and peers, the estimate reported to the IRB was 30 minutes. Based on collected data, most participants were able to complete the survey within 20 minutes. The survey begins with obtaining consent and a brief explanation of the study, after which the participants will complete the measures outlined above.

Once participants complete the outlined measures, the final page will have a link redirecting to an external website to disburse incentives. On this page, participants are also offered a set of roughly a dozen online resources related to potential concerns about disordered eating or broader mental health concerns. All participants who choose to enter their email have the option to enter a drawing for one of three $50 Amazon gift cards. IP address collection was manually disabled in Qualtrics, ensuring that there is no way to link a specific set of responses to any of the emails that were submitted in the incentive disbursement survey. Initially, we had hoped to include a summary of the participants’ responses at the end of the survey as well to promote transparency and to incentivize participation from people who may be curious about their personal scores. We were advised by the IRB not to do so, although if we had a longer timeframe, we would have liked to discuss this option further.

Hypotheses
The central hypothesis of this study is that those who are Asian by race will report greater eating disturbances as measured by the EDE-Q than participants who do not identify as Asian. Asians will most likely also report more negative attitudes toward seeking professional help as indicated by the ATSPPH-SF. We expect that Asian males compared to non-Asian males will have a greater difference in EDE-Q scores than between the corresponding female groups. Additionally, we expect that participants who adhere more strongly to Asian values as determined by the AAVS and are more perfectionistic as measured by the FMPS will also report higher EDE-Q scores. In order to determine whether the EDE-Q even appropriately measures eating psychopathology in Asians as well as it does for other race groups, we conducted differential item functioning (DIF) analyses to determine whether or not the measure is biased with respect to those characteristics. Along those lines, we also conducted a factor analysis (EFA) of the EDE-Q, one of the most commonly used measures of disordered eating, to determine how the factor structure holds for Asian Americans, examining whether extracting four factors would resemble the published scale content.

Results
Tables

Table 1. Sociodemographic Characteristics of Participants (N = 297).

Table 2. Means, Standard Deviations, and One-Way Analyses of Variance in ACIEA Measures (N = 297). *p<.05,  ***p < .001. Note. EDE-Q = Eating Disorders Examination Questionnaire; BMI = Body Mass Index; SL-ASIA = Suinn-Lew Asian Self-Identity Acculturation Scale; AAVS-M = Asian-American Values Scale (Multidimensional); FMPS = Frost Multidimensional Perfectionism Scale; ATSPPH-SF = Attitudes Towards Seeking Professional Psychological Help – Short Form.

Table 3. Regression Analysis Using Perfectionism Variables to Predict EDE-Q Global Score. Note. Constant=-0.86, F(4, 294)=20.30, p<.0005, R2=.22

Table 4. Regression Analysis Using Collectivism Variables to Predict EDE-Q Global Score. Note. Constant=1.671, F(4, 304)= 7.20, p<.0005, R2=.09

Table 5. Regression Analysis Using Conformity Variables to Predict EDE-Q Global Score. Note. Constant=2.11, F(4, 304)=5.31, p<.0005, R2=.07

Table 6. Results from Exploratory Factor Analyses of the Eating Disorders Examination Questionnaire (EDE-Q) Using Principal Axis Factoring and PROMAX Rotation, Extracting Four Factors. Note. SW = shape or weight.

Table 7. Inter-Factor Correlations from EDE-Q Based on PROMAX Rotation of Four Factor PAF Extraction.

Demographics
As seen in Table 1, 141 Asian and 156 non-Asian participants completed the ACIEA survey. The mean age of the sample was 29.5 years (SD=13.5) with a significantly positively skewed distribution and a mode of 20 years.

Primary Aim: Testing Asian versus Non-Asian Differences in Disordered Eating Behavior
Our central hypothesis was that Asians would report higher rates of disordered eating behavior as measured by the EDE-Q, which in turn we predicted would correlate with higher scores on the values of collectivism, conformity to norms, and perfectionism. Table 2 presents the mean scores and results of independent-samples t-tests for Asian and non-Asian participants for all measures. Asian participations reported significantly lower scores (p<.001) for all subscales of the EDE-Q as well as the Global Score. While the SL-ASIA is not linear in nature, Asian participants were more likely to end up with the scores indicating Asian identification as expected. The collectivism subscale of the AAVS, which was expected to be the most correlated with identifying as Asian, was actually not significantly different between Asian and non-Asian participants (p=.995). Conformity to norms (p=.002), emotional self-control (p<.001), and the total score (p<.001) were significantly higher for Asian participants than non-Asian. The only subscale of the FMPS that was significantly different between the groups was parental expectations and criticism, for which Asians reported a higher score (p<.001). Lastly, Asian participants reported more positive attitudes towards seeking professional help based on ATSPPH scores (p=.015).

We hypothesized that being Asian would additionally change the effect of gender on EDE-Q scores. To determine whether the effect of gender on disordered eating as a function of being Asian, we conducted a 2 (gender: male, female) X 2 (race: Asian, non-Asian) analysis of variance (ANOVA); see Figure 1. We found a significant main effect of being Asian on EDE-Q scores, such that those who were Asian had lower EDE-Q scores (M = 1.63, SD = 1.22) than those not in college (M =2.26, SD = 1.57), F (1, 333) = 10.03, p = .002, partial η2 = .029. We did not find a significant main effect of gender on EDE-Q scores such that those who were male had a similar level of disordered eating (M = 1.56, SD = 1.27) as those who were female (M = 2.04, SD = 1.46), F (1, 333) = 3.70, p = .055, partial η2 = .011. Finally, there was no significant interaction between gender and being Asian on EDE-Q scores, F (1, 333) = .07, p = .792, partial η2 = .000.

To determine if perfectionism, collectivism, conformity to norms, and their interactions with being Asian were significant predictors to EDE-Q scores, we conducted three regression analyses. Table 3 shows that perfectionism as measured by the FMPS total score did is in fact a significant predictor of EDE-Q Global Scores, b=.08, SE=.01, t(294)=6.38, p=.000. Additionally, the interaction between perfectionism and Asian status was significant, b= -.01, SE=.01, t(294)= -1.97, p=.050. An R2 of .22 for this model, which also included gender and Asian status as independent predictors that were not significant, suggests that 22% of the variability observed in EDE-Q Global Scores can be explained by the combination of perfectionism, gender, Asian status, and the interaction of perfectionism and Asian status (p<.001). Similar models were constructed for collectivism and conformity to norms, as displayed in Tables 4 and 5; but these variables and their interactions with being Asian were not found to be significant. The overall trends of these relationships using perfectionism, collectivism, and conformity are displayed graphically in Figures 3, 4, and 5, respectively. Figure 3 illustrates that the positive relationship between perfectionism and EDE-Q scores is weaker for Asian than non-Asian participants. Additionally, the parallel lines demonstrate the observed gender differences in scores, such that males across the board had lower predicted EDE-Q scores than women. These parallel lines by gender are maintained Figures 4 and 5. In Figure 4, the slope of the lines suggests that collectivism and EDE-Q have a stronger positive relationship for Asian than non-Asian participants, although this relationship was not found to be significant as shown in Table 4. Figure 5 suggests a possibly negative relationship between conformity to norms and EDE-Q scores for non-Asians, but a positive relationship for Asians. However, these relationships again were not found to be significant (Table 5).

Secondary Aim 1: Examining Reproducibility of Fairburn’s Published EDE-Q Structure
To examine whether Fairburn’s suggested four-factor structure of the EDE-Q as defined by the four subscales (Fairburn & Beglin, 1994) holds similarly for Asian and non-Asian populations, we conducted an exploratory factor analysis split by Asian and non-Asian participants, extracting four factors to examine whether they resembled the published scale content as well as each other. Table 6 shows the item loadings for a four-factor solution using principal axis factoring (PAF) and PROMAX rotation, with correlation for the four identified factors shown in Table 7. With the exception of most items from the Shape Concern and Weight Concern subscales loading onto the same factor, the data suggests a factor structure different from the four subscales established by Fairburn. However, most items loaded onto the same factor between the Asian and non-Asian samples, suggesting those differences are replicated for these two groups.

Secondary Aim 2: Examining Differential Item Functioning (DIF) between Asian and Non-Asian Respondents
We also conducted differential item functioning (DIF) analyses to determine whether items of the EDE-Q predicted disordered eating symptom severity differently for Asians vs non-Asian participants. These analyses used a regression-based approach, where each item was predicted by the sum of the rest of the items (excluding the item in question), as well as a dummy code for Asian identity, and an interaction term for Asian status * trait score. A significant coefficient for the trait score predicting the item indicates that the item is a valid marker for the construct. A significant effect for Asian status would mean that there were group differences in the behavior for reasons besides trait disordered eating. A significant interaction would indicate that the behavior coded by the item was more strongly related to one group than the other.

Analyses of 22 items included in calculations of the global scale and subscales yielded no indication of uniform DIF, while items #8 (p=.002), #19 (p=.006), and #21 (p=.002) appear to have significant non-uniform DIF (Figure 2). An alpha level of 0.01 was used to determine statistical significance. For all analyses, there was a statistically significant positive relationship between item score and EDE-Q Global Score, indicating that each item contained construct-relevant variance. For items #8 (thinking about shape or weight has made it difficult to concentrate), #19 (eating in secret), and #21(concern about others seeing you eat), higher Global Scores in Asian participants (indicating more disordered eating behavior) became less predictive of their increase in item scores. The only observable exception was item #28 (discomfort around others seeing your shape or figure), which seemed to have the opposite effect, as Global scores increase for Asians, although this model was marginally significant with p=.015.

Discussion
The goal of this study was to determine whether Asian Americans are more or less at risk of engaging in disordered eating and therefore of developing EDs than the rest of the population. Based on prior research, we hypothesized that Asians would report more disordered eating as measured by the EDE-Q which would be related to having higher levels of the risk factors of perfectionism (FMPS), collectivism, and conformity to norms (AAVS). We also expected that there would be DIF of the EDE-Q as a result of culturally differentiated relevance of certain items.

The data seem to disprove most of our initial hypotheses. Remarkably, Asians actually reported less disordered eating behaviors as measured by the EDE-Q than non-Asians, supporting literature suggesting being Asian has no impact or may even be a protective factor against developing EDs. This was coupled with the observation that Asians still had significantly lower BMIs, suggesting that having a lower BMI does not necessarily indicate increased risk for EDs. Another surprising finding was that collectivism, commonly regarded as a cultural value that differentiates Asian from Western societies, was not significantly different between participants who were Asian and not. It may be that participants who were interested in a survey about cultural differences were more likely to share a collectivistic attitude, or Asian Americans may not be as collectivistic as people in their countries of origin.

The fact that the items for Shape Concern and Weight Concern largely appeared to load onto the same factor has also been supported by more recent research into the factor structure of the EDE-Q, with some groups suggesting that this should be regarded as one factor and that there should be a separate appearance factor (Phillips et al., 2018). This was also a trend observed for both Asian and non-Asian participants. Overall, there is no evidence that the factor structure of the EDE-Q would be significantly different between Asian and non-Asian populations.

The lack of uniform DIF implies that there are no items on the EDE-Q that indicate a direct impact on the relationship between item score and overall score based on being Asian alone. However, the non-uniform DIF identified in items 8, 19, 21, and possibly 28 imply that Asians who exhibit more disordered eating may not demonstrate certain behaviors in the same way or to the same degree as those who are not Asian. In other words, there is a significant interaction effect of being Asian and having a higher EDE-Q score that warrants further exploration.

Bias
Both sampling bias and non-response bias played a significant role in the demographic distribution of the sample. It is typical of college-based research to see an overrepresentation of college-aged, college-educated participants as we observed in this study. Additional sources of bias are most clearly illustrated in the use of ResearchMatch, the most common source of participants. The two decision points on the part of the participants - indicating whether they are interested in finding out more about the study and later actually accessing the link to complete it - gave people two opportunities to not participate, whether out of lack of interest or other reasons. The types of people who did proceed at both points were likely people who were interested in the research topics of mental health, disordered eating, or cultural differences. This is reflected in the demographic breakdown of Table 1. College students, especially psychology students, tend to skew female. Women are also more likely to be interested in the topic of disordered eating. Over half the sample had some history of seeking professional help for a mental health concern, and 16% have received a formal ED diagnosis, nearly 8 times higher than the typical rate observed in the population. More funding and recruiting manpower for sophisticated and nuanced recruitment to gather a more representative sample could alleviate these concerns in future research. Additionally, further analyses conducted on the dramatic difference in the proportion of Asians (3%) vs non-Asians (28%) who reported having had an ED diagnosis could have revealed whether the comparably low scores on the EDE-Q in Asians compared to non-Asians was more attributable to a lower proportion of participants with ED history than to being Asian. On the other hand, the participation patterns functionally oversampled cases with concerns about eating disorders and Asians, two groups that have in the past been difficult to recruit and often are under-represented in research. The groups were large enough to provide adequate statistical power for primary analyses, and strong tests of DIF, as a result.

Future directions
The time constraints set by the single school-year scale of an undergraduate honors thesis limited the scope of this study. Research for Me is an exceptionally promising recruitment platform for conducting nationwide studies that could better reflect population-level trends than conducting surveys locally in Chapel Hill. More time would have allowed us to explore setting more advanced demographic filters, which go as far as filtering by health diagnosis and medication. Another topic we had wished to explore more thoroughly was the idea of providing results summaries to participants. While we attempted to include language that clearly states that survey results would not take the place of a professional diagnosis in several places on the initial survey with the knowledge that there is no shortage of surveys and quizzes freely available on the internet that have not been validated at all, for the sake of time we could not engage in a potentially productive discussion with the IRB to see if we could balance transparency and safety in delivering any results to participants in a beneficial way. We hope that in the future, research becomes more transparent in the sharing of results, especially as the pace and availability of technology progresses. Knowledge does not always need to be regarded as a liability, and if communicated correctly, it can provide participants with a sense of empowerment and even curiosity to learn more about psychology and mental health issues.

More attention should be directed to ensuring that psychological measures retain psychometric properties across populations. While there was no evidence of uniform DIF in the EDE-Q for Asian vs non-Asian participants, there were several items that indicated non-uniform DIF. If measures, especially those with clinical purposes, are developed without taking diverse populations into account, they will underserve those populations which could negatively influence the treatment they receive. Especially for disorders as serious as EDs, accurate and equitable assessment may be a matter of life and death at a population level. Within the framework of minority stress theories, we would also expect that those who have low or moderate Western acculturation as measured on the SL-ASIA might take social norms into heavier consideration and thus experience greater distress around body image, with higher scores on the EDE-Q as well. Luckily, with growing globalization, more research groups are looking to develop cross-cultural adaptations of existing measures or otherwise develop measures that are more widely applicable at a global scale, which suggests that the field is moving in the right direction. At the end of the day, marginalized populations, whether that mean racial/ethnic minorities, sexual/gender minorities, immigrants, or low SES, are often the least likely to be represented in psychological research, least likely to access adequate treatment, and most likely to be at risk for various disorders. In order for that to change, researchers in every subset of psychology must make an effort to be inclusive of these groups while recognizing that there may be unique challenges that need to be approached differently.

Conclusions
The goal of the study was to examine the similarities and differences between Asian and non-Asian participants with respect to disordered eating and its associated risk factors, with a focus on the cultural and social attributes of Asian Americans. We had hypothesized that Asian Americans would exhibit more perfectionism, collectivism, and conformity to norms and in turn more disordered eating. We found that this was not the case, and that Asian participants actually showed less disordered eating than non-Asian participants. While Asians did show more perfectionism related to parental expectations and criticism as well as conformity to norms, they did not show significantly different collectivism. Surprisingly, Asian participants also exhibited more positive attitudes towards seeking professional psychological help, which goes against previous literature about mental health stigma in this population. While the data supported a different factor structure for the EDE-Q from the one defined by the original authors, there were no significant differences between Asian and non-Asian participants. Additionally, there was evidence of non-uniform DIF for some items, but no uniform-DIF that would have suggested bias in the measure. More research is needed to determine if Asian Americans at different levels of acculturation are at higher risk of disordered eating, and if they display more of the risk factors mentioned previously. Lastly, more research should be conducted on disordered eating in males with specific attention to males across races and cultures, in addition to further research into disordered eating and EDs for minority groups more broadly.

Figures
Figure 1

EDE-Q scores as a function of gender and Asian status

Note. Asians had lower average EDE-Q scores, F (1, 336) = 19.57, p = .002. Women (in red) had higher scores, F (1, 336) = 3.70, p = .055. There was no significant interaction.

Figure 2

Graphs of predicted values of the EDE-Q Global Score by Item for Items 8, 19, 21, and 28

Note. These figures show the four items that had potentially statistically significant DIF. None showed intercept bias. Three showed evidence of non-uniform DIF, with weaker association between the item and the construct of disordered eating behaviors in Asians. One showed the opposite, with a steeper association between discomfort towards others seeing shape or figure and eating behavior problems.

Figure 3

Plot of Unstandardized Predicted Value of EDE-Q Based on Regression Using Perfectionism

Note. Red = Asian. The appearance of double lines is due to the inclusion of sex in the regression model. Men tended to have lower scores, so the lower line (with fewer cases) is the male trend line for each group.

Figure 4

Plot of Unstandardized Predicted Value of EDE-Q Based on Regression Using Collectivism

Note. Red = Asian. The appearance of double lines is due to the inclusion of sex in the regression model. Men tended to have lower scores, so the lower line (with fewer cases) is the male trend line for each group.

Figure 5

Plot of Unstandardized Predicted Value of EDE-Q Based on Regression Using Conformity

Note. Red = Asian. The appearance of double lines is due to the inclusion of sex in the regression model. Men tended to have lower scores, so the lower line (with fewer cases) is the male trend line for each group.